AMD Reverses Their Blunder - Game Support Returns to RDNA 1/2

[-]

EiffelPower76@reddit

Funny how these youtubers surf on the useless (temporary) shistorms on Internet

[-]

Culbrelai@reddit

AMD did NOT reverse their blunder. AMD unboxed is at it again, spinning the corporate spin from AMD. Bravo.

They did not say directly they are not in maintenance mode. They are keeping the driver branches seperate. Its such a crock of shit lmao

[-]

zsaleeba@reddit

Separate branches is fine, and quite likely makes sense from a software development perspective if the newer architectures have diverged a lot from the earlier ones.

What I'm still not convinced about is that they'll apply the same amount of effort optimising new games on the old architectures that they will on the new ones. And notice that they didn't say they'll optimise all the new games, just that they'll "optimise new games". I wouldn't be suprised if RDNA 1/2 get less attention than 3/4 in future.

[-]

_hlvnhlv@reddit

It doesn't make any sense at all.

Just look at Nvidia's driver, or Mesa (linux), which is a massive "driver package" that supports almost every thing released in the last 20 years.

In fact, Mesa / RADV has implemented RT support on Vega +, and you can play full blown RT games on it at playable framerates.

This is just AMD trying to fuck us over. Remember the debacle with the 16MB bioses and Zen3? Yeah

[-]

master_assclown@reddit

AMD is splitting up the driver branches because RDNA 1 and 2 were not AI focused and lack dedicated hardware for newer, more AI focused/efficient data sets. With the move toward AI and features that require AI/similar data sets, it makes perfect sense to split the driver branches, separating RDNA 1/2 from the rest. RDNA 3 will be next because it's limited in the data sets that it supports so by the time UDNA/RDNA 5 hits shelves or around the same lifespan that RDNA 4 has at this time, RDNA 3 very well could get some sort of separation as well.

AMD was just very clearly way behind the ball on the AI shift/boom. The same can be said about Nvidia and where their driver branches separate. GPUs with Tensor Cores are on one branch while GPUs without them are on another.

[-]

zsaleeba@reddit

FYI I'm a software developer and I've worked on teams where we've split our source between models, for exactly this reason. It does make some sense, although there are also downsides.

[-]

Merdiso@reddit

Obviously they don't, that's the disadvantage when you always change your architecture significantly because you didn't design it with 'forward thinking' in mind in the first place - e.g. Blackwell (2014) is still based on Turing (2018) as a core philosophy (Shaders/RT/Tensor Cores), now look at RDNA, not even RDNA4 does have the same principle, but wait, RDNA5 finally will, but guess what will happen with RDNA 3/4 next.

[-]

Ecstatic_Secretary21@reddit

Will people actually praise AMD for the u-turn?

Now I am more sure they will silently turn off features instead of publicly

[-]

noiserr@reddit

People still criticize AMD for not supporting some motherboards with Zen3 initially.

Literally calling AMD unethical because of it, even though AMD listened and all those motherboards ended up being supported.

This space is so toxic and anti consumer it's ridiculous. Nvidia has been using their monopoly for anti-comsumer vendor lockins the entire time and no one talks about it.

[-]

_hlvnhlv@reddit

Is there something to praise? They haven't done jack shit, no new DX or Vulkan support, nor FSR.

[-]

According_Spare7788@reddit

What is there to praise?? That they managed to somewhat save their ass over what is at best, a disastrous pr incident over nothing?

[-]

constantlymat@reddit

It is honestly discouraging how many people embraced AMD's cheap PR ploy even though nothing substantive has changed.

I also blame Tomshardware and the ones copying them for just claiming AMD reversed course when they didn't

[-]

Culbrelai@reddit

This exactly. They are using corporate platitudes. And it was super effective. I’ll believe it when I see game optimizations for RDNA 2 at the same time as later ones.

[-]

jenny_905@reddit

There was no reversal

[-]

HippoLover85@reddit

I don't think amd ever intended to not support these gpus. They have a strong track record of providing decent support for old gpus.

Hell vega 64 still works on bf6 day 1.

I think they just suck at pr and communicate, which has always been tried and true.

[-]

a2e5@reddit

Yeah. Not doing "game-ready" profiles really isn't the same as "we won't guarantee that the render APIs used by games work any more", but ehhhhhhh nuances die on the Internet.

[-]

skinlo@reddit

The internet just likes getting angry, it turns into a circlejerk.

[-]

ShogoXT@reddit

Please don't let go about the Vulkan instructions! That will have huge application and usage ramifications in the future!

[-]

Nordmuth@reddit

No, AMD certainly did not reverse course. The new blog statement as far I can parse is just same thing as previously, this time worded by someone vaguely competent.

Game support sounds great, until you realize it can mean whatever AMD wants to mean. Game runs three months after launch? Game support! Basically same thing as market needs, reduced support but this time with nicer phrasing.

Presumably no more RDNA2 game optimization, unlikely inclusion of INT8 FSR4, which runs even on Windows with driver version 23.9.1 via Optiscaler. Compared to the GTX 1000 series nominally having game optimizations up until this October, and RTX 2000 series having received DLSS4? The end result is inferior support by a company that has a weaker product and market position.

This is less impactful to me than some others, as I am moving over to Linux as much as possible as the Windows 10 ESU period winds down, but 2020 products that are as popular as RDNA2 do not belong in a maintenance/legacy branch in 2025, let alone 2025 products.

I can only assume that AMD is slowly exiting consumer GPU space at this point? Otherwise they need to do some heavy handed trimming in their Radeon division leadership.

[-]

DadSchoorse@reddit

People shouldn't care about these game optimizations that nobody (without internal access) can really quantify.

What we can quantify is Vulkan/D3D12 feature support. Features that do not require new hardware. And there, Vega and older have no received anything since they were moved to the "maintenance" branch. And the recent driver release only adds VK_KHR_shader_untyped_pointers for RDNA3+, despite this being a quality of life extension that all hardware can support.

This means either graphics programmers have to plan around the RNDA2 drivers for the next 3 years and not require anything new, or they only support RDNA3+. Both are bad, and this is only caused by AMD wanting to save money in the wrong place.

[-]

uzzi38@reddit

Vega is treated as "legacy", not "maintainance" like RDNA1 and RDNA2.

[-]

werpu@reddit

Well they saw they made a mistake and reverted. So they are listening!

Not bad!

[-]

Merdiso@reddit

This doesn't sound as a 'revert' per se, which is quite terrible, because it tells me FSR4 will not be released on RDNA2 whatsoever, although it could technically run it, but on the other hand, how many times do people need to realize AMD simply isn't up to par with nVIDIA when it comes to driver support and forward architecture thinking?

Turing (2018) has dedicated RT/Tensor cores specifically designed for RT/DLSS, whereas RDNA3 (2022 - 4 YEARS LATER) doesn't, and with that, I think I said it all.

In fact, I'm already sure RDNA 3 and even 4 will have the same faith, as UDNA/RDNA 5 will be a totally different architecture - which, by the way, will be literally a copy of Turing from a core perspective, which was released, once again, in 2018 -- 'Radiance Cores' are the ray-tracing cores, and 'Neural Arrays' are the tensor cores - obviously, much more powerful than Turing's, but you get the point - AMD is about 9 years behind nVIDIA in terms of architecture philosophy at this point.

So do not act surprised that 5700 XT didn't have more value compared to 2070S at 100$ more as some of people thought, there's a reason "nVIDIA -50$" doesn't work anymore, those 50$ will bite you in your ass next time and people have woken up.

[-]

loozerr@reddit

Old gpus like what rog ally is shipping with.

[-]

Firefox72@reddit

One has to be a profesional HUB hater to interpret this situation like this

Their pressure and very loud video no doubt contributed to this.